Linan Yue

Dr. Linan Yue


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 July 2023
Verified
Conference Paper
Open access
WGWeibo Gao
Weibo Gao
Hao WangHao Wang
Hao Wang
QLQi Liu
Qi Liu
FWFei Wang
Fei Wang
XLXin Lin
Xin Lin
SWShijin Wang
Shijin Wang
Cognitive diagnosis (CD) aims to reveal the proficiency of students on specific knowledge concepts and traits of test exercises (e.g., difficulty). It plays a critical role in intelligent education systems by supporting personalized learning guidance. However, recent developments in CD mostly concentrate on improving the accuracy of diagnostic results and often overlook the important and practical task: domain-level zero-shot cognitive diagnosis (DZCD). The primary challenge of DZCD is the deficiency of student behavior data in the target domain due to the absence of student-exercise interactions or unavailability of exercising records for training purposes. To tackle the cold-start issue, we propose a two-stage solution named TechCD (Transferable knowledgE Concept grapH embedding framework for Cognitive Diagnosis). The fundamental notion involves utilizing a pedagogical knowledge concept graph (KCG) as a mediator to connect disparate domains, allowing the transmission of student cognitive signals from established domains to the zero-shot cold-start domain. Specifically, a naive yet effective graph convolutional network (GCN) with the bottom-layer discarding operation is initially employed over the KCG to learn transferable student cognitive states and domain-specific exercise traits. Moreover, we give three implementations of the general TechCD framework following the typical cognitive diagnosis solutions. Finally, extensive experiments on real-world datasets not only prove that Tech can effectively perform zero-shot diagnosis, but also give some popular applications such as exercise recommendation.

24 March 2024 in Proceedings of the AAAI Conference on Artificial Intelligence
Verified
Journal Article
Open access
PDF file
WGWeibo Gao
Weibo Gao
QLQi Liu
Qi Liu
Hao WangHao Wang
Hao Wang
LYLinan Yue
Linan Yue
HBHaoyang Bi
Haoyang Bi
YHYuanjing He
Yuanjing He
Cognitive diagnosis seeks to estimate the cognitive states of students by exploring their logged practice quiz data. It plays a pivotal role in personalized learning guidance within intelligent education systems. In this paper, we focus on an important, practical, yet often underexplored task: domain-level zero-shot cognitive diagnosis (DZCD), which arises due to the absence of student practice logs in newly launched domains. Recent cross-domain diagnostic models have been demonstrated to be a promising strategy for DZCD. These methods primarily focus on how to transfer student states across domains. However, they might inadvertently incorporate non-transferable information into student representations, thereby limiting the efficacy of knowledge transfer. To tackle this, we propose Zero-1-to-3, a domain-level zero-shot cognitive diagnosis framework via one batch of early-bird students towards three diagnostic objectives. Our approach initiates with pre-training a diagnosis model with dual regularizers, which decouples student states into domain-shared and domain-specific parts. The shared cognitive signals can be transferred to the target domain, enriching the cognitive priors for the new domain, which ensures the cognitive state propagation objective. Subsequently, we devise a strategy to generate simulated practice logs for cold-start students through analyzing the behavioral patterns from early-bird students, fulfilling the domain-adaption goal. Consequently, we refine the cognitive states of cold-start students as diagnostic outcomes via virtual data, aligning with the diagnosis-oriented goal. Finally, extensive experiments on six real-world datasets highlight the efficacy of our model for DZCD and its practical application in question recommendation. The code is publicly available at https://github.com/bigdata-ustc/Zero-1-to-3.

13 December 2024
Verified
Book Chapter
Closed access
PDF file
LYLinan Yue
Linan Yue
QLQi Liu
Qi Liu
YDYichao Du
Yichao Du
WGWeibo Gao
Weibo Gao
YLYe Liu
Ye Liu
FYFangzhou Yao
Fangzhou Yao

1 January 2023
Verified
Conference Paper
Open access
PDF file
LYLinan Yue
Linan Yue
QLQi Liu
Qi Liu
LWLi Wang
Li Wang
YAYanqing An
Yanqing An
YDYichao Du
Yichao Du
ZHZhenya Huang
Zhenya Huang
Linan Yue, Qi Liu, Li Wang, Yanqing An, Yichao Du, Zhenya Huang. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.

11 April 2024 in IEEE Transactions on Knowledge and Data Engineering
Verified
Journal Article
Closed access
LYLinan Yue
Linan Yue
QLQi Liu
Qi Liu
BJBinbin Jin
Binbin Jin
HWHan Wu
Han Wu
YAYanqing An
Yanqing An
Massive legal documents have promoted the application of legal intelligence. Among them, Legal Judgment Prediction (LJP) has emerged as a critical task, garnering significant attention. LJP aims to predict judgment results for multiple subtasks, including charges, law articles, and terms of penalty. Existing studies primarily focus on utilizing the entire factual description to produce judgment results, overlooking the practical judicial scenario where judges consider various crime circumstances to decide verdicts and sentencing. To this end, in this article, we propose a circumstance-aware LJP framework (i.e., NeurJudge) by exploring the circumstances of crime. Specifically, NeurJudge first separates the factual description into different circumstances with the predicted results of intermediate subtasks and then employs them to yield results of other subtasks. Besides, as confusing verdicts may degrade the performance of LJP, we further develop a variant of NeurJudge (NeurJudge+) that incorporates the semantics of labels (charges and law articles) into facts to yield more expressive and distinguishable fact representations. Finally, to provide explanations for LJP, we extend NeurJudge to an explainable LJP framework E-NeurJudge with a cooperative teacher-student system. The teacher system is NeurJudge which exploits legal particularities well but lacks explanation capability. The student system is a rationalization method that provides explainability but fails to utilize legal particularities. To combine the advantages of the above methods, we use a transferring function to transfer legal particularities from the teacher to the student, making a trade-off between yielding LJP results and rendering them explainable. Extensive experimental results on real-world datasets validate the effectiveness of our proposed frameworks.

13 May 2024
Verified
Conference Paper
Open access
LYLinan Yue
Linan Yue
QLQi Liu
Qi Liu
YLYe Liu
Ye Liu
WGWeibo Gao
Weibo Gao
FYFangzhou Yao
Fangzhou Yao
WLWenfeng Li
Wenfeng Li
Graph Neural Networks (GNNs) have achieved impressive results in graph classification tasks, but they struggle to generalize effectively when faced with out-of-distribution (OOD) data. Several approaches have been proposed to address this problem. Among them, one solution is to diversify training distributions in vanilla classification by modifying the data environment, yet accessing the environment information is complex. Besides, another promising approach involves rationalization, extracting invariant rationales for predictions. However, extracting rationales is difficult due to limited learning signals, resulting in less accurate rationales and diminished predictions. To address these challenges, in this paper, we propose a Cooperative Classification and Rationalization (C2R) method, consisting of theclassification and therationalization module. Specifically, we first assume that multiple environments are available in theclassification module. Then, we introduce diverse training distributions using an environment-conditional generative network, enabling robust graph representations. Meanwhile, therationalization module employs a separator to identify relevant rationale subgraphs while the remaining non-rationale subgraphs are de-correlated with labels. Next, we align graph representations from theclassification module with rationale subgraph representations using the knowledge distillation methods, enhancing the learning signal for rationales. Finally, we infer multiple environments by gathering non-rationale representations and incorporate them into theclassification module for cooperative learning. Extensive experimental results on both benchmarks and synthetic datasets demonstrate the effectiveness of C2R. Code is available at https://github.com/yuelinan/Codes-of-C2R.